W3C home > Mailing lists > Public > www-webont-wg@w3.org > September 2002

RE: semantics document revised

From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
Date: Thu, 12 Sep 2002 17:49:12 +0200
To: "pat hayes" <phayes@ai.uwf.edu>, <www-webont-wg@w3.org>
Message-ID: <BHEGLCKMOHGLGNOKPGHDOECACAAA.jjc@hpl.hp.com>

Pat:
> There is a new version of the OWL/RDF semantics document at
>
> http://www.coginst.uwf.edu/users/phayes/RDFS2OWL-G.html

I like Pat's doc.

Peter has highlighted a few bugs to be fixed though, two of which I find
interesting and may merit some discussion.

Peter:
>- The conditions on rdfs:domain, rdfs:range, rdfs:subClassOf, and
>  rdfs:subPropertyOf are too weak.

Peter (in detail)
> Fixes [sic] to RDFS.
>
> <x,y> in IEXT(IS(rdfs:subClassOf)) iff ICEXT(x) <= ICEXT(y)
> <x,y> in IEXT(IS(rdfs:subPropertyOf)) iff IEXT(x) <= IEXT(y)
> <x,y> in IEXT(IS(rdfs:domain)) iff <z,w> in IEXT(x) implies z in ICEXT(y)
> <x,y> in IEXT(IS(rdfs:range)) iff <w,z> in IEXT(x) implies z in ICEXT(y)

These rules seem to belong to the strong semantics of OWL (Pat's section 4).
It is intersting that they do not mention OWL.
They *could* be restricted to owl:Class for the purposes of the strong
semantics.

> - Restrictions have to be relative to owl:Thing.
This seems correct, and a shame.
There would be problem if for instance rdfs:Class was a member of a
restriction. But by the rules Pat gives it is likely to be a member of some
restriction (e.g. one with cardinality=0 on some property that cannot apply
to classes). Section 3 does a valiant job of capturing the essential
semantics of OWL without the comprehension principle (and hence with fewer
closure rules) over the whole of the RDF space. Personally I don't see an
easy way to keep the treatment of restrictions within this philosophy (Pat
is, of course, a master magician; so I am expecting a pleasant surprise).
(I could expand my example if it would help)

> - The definition of sameXXAs, and some other OWL constructs, should be
iff,
>   perhaps restricted to owl:Thing.  If not restricted to owl:Thing, then
>   the domains and ranges of the properties have to be removed.

I would have thought the easiest editorial fix to these bugs is to
copy-paste the table of "if then" rules from section 3 to section 4 and
reverse the implications.



> - There are many divergences between the constructs described in the
>   document and the constructs of OWL.

A couple of minor bugs I picked up were teh occasional use of owl:List
vocabulary instead of rdf:List vocab, and the use of owl:UniqueProperty (for
datatype properties?) as well as owl:FunctionalProperty. I think we are
using owl:FunctionalProperty for both.
Also

owl:minCardinality rdfs:range owl:DataTypeProperty .

should read

owl:minCardinality rdfs:range owl:DataTypeValue .
Received on Thursday, 12 September 2002 11:49:34 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:52 GMT